Adaptive Multi-level Feature Fusion for Scene Ancient Chinese Text Recognition

TU Chaohu; YI Yaohua; WANG Kaili; PENG Jibing; YIN Aiguo

doi:10.13203/j.whugis20230176

您当前的位置：

首页 >

文章列表页 >

Adaptive Multi-level Feature Fusion for Scene Ancient Chinese Text Recognition

Cartography and GIS | 更新时间：2026-11-05

- Adaptive Multi-level Feature Fusion for Scene Ancient Chinese Text Recognition
- Geomatics and Information Science of Wuhan University Vol. 50, Issue 12, Pages: 2592-2600(2025)
- 作者机构：
  
  1.武汉大学遥感信息工程学院,湖北武汉,430079
  2.武汉大学数字成像与智能感知研究中心,湖北武汉,430079
  3.武汉工程大学计算机科学与工程学院,湖北武汉,430205
  4.珠海奔图电子有限公司,广东珠海,519060
- 作者简介：
- 基金信息：
- DOI：10.13203/j.whugis20230176
  CLC： P208
- Received：26 October 2024，
  
  Published：05 December 2025
- 稿件说明：
移动端阅览
涂超虎,易尧华,王凯丽，等.自适应多级特征融合的场景古汉字识别[J].武汉大学学报（信息科学版）,2025,50(12):2592-2600.

TU Chaohu,YI Yaohua,WANG Kaili,et al.Adaptive Multi-level Feature Fusion for Scene Ancient Chinese Text Recognition[J].Geomatics and Information Science of Wuhan University,2025,50(12):2592-2600.
涂超虎,易尧华,王凯丽，等.自适应多级特征融合的场景古汉字识别[J].武汉大学学报（信息科学版）,2025,50(12):2592-2600. DOI：10.13203/j.whugis20230176

TU Chaohu,YI Yaohua,WANG Kaili,et al.Adaptive Multi-level Feature Fusion for Scene Ancient Chinese Text Recognition[J].Geomatics and Information Science of Wuhan University,2025,50(12):2592-2600. DOI：10.13203/j.whugis20230176

摘要

自然场景中的古汉字图像具有背景复杂、字符数量庞大、书体形式多样的特点，其字符与书体形式众多导致了文本结构复杂度不同，现有研究方法未针对性解决复杂结构古汉字的识别难题。针对这一问题，提出一种自适应多级特征融合网络，首先根据古汉字的结构复杂度，自适应选择融合古汉字浅层细节信息和高层语义信息，获取古汉字的高区分度特征，提高模型对古汉字的识别能力。然后使用最大边界余弦损失，增大古汉字的类间间距，提高模型对相似结构古汉字特征的判别能力。实验结果表明，所提方法在多场景古汉字数据集上Top1识别准确率为79.58%，与目前最优的基于特征对齐和交叉域融合的方法相比提高了3.27%，提高了场景古汉字的识别准确率。

Abstract

Objectives

Ancient Chinese text are widely distributed in inscriptions

couplets

stone engravings and other scenes

which have the characteristics of complex background

large number of characters

and diverse writing forms. The large number of characters and writing forms directly lead to the difference in text structure complexity.

Methods

To solve the difficulty of recognizing ancient Chinese text with complex structures

we propose an adaptive multilevel feature fusion network. First

ResNet152 is the main backbone network

and its deeper network and residual structure can fit more parameters to learn the features of ancient Chinese text and avoid the degradation of the model. Second

according to the structural complexity of ancient Chinese text

the importance of each feature map is automatically obtained through learning

so that the model adaptively selects and merges the shallow detail information and high-level semantic information of ancient Chinese text

obtains the high discrimi-nation features of ancient Chinese text and improves the recognition ability of the model. Finally

the maximum boundary cosine loss is used to minimize the cosine similarity between different ancient Chinese text

increase the inter-class distance of ancient Chinese text

and reduce the intra-class distance between similar Chinese text. Combined with the cross entropy loss function as a loss function

the proposed model can improve the discrimination ability of ancient Chinese text with similar structures.

Results

The experimental results show that when the multistage feature fusion module is added to the proposed method

the Top1 accuracy rate is increased by 1.59%

and when the maximum boundary cosine loss function is added

the Top1 accuracy rate is increased by 1.09%. The best effect of Top1 identification accuracy rate on the multi-scene ancient Chinese character dataset is 79.58%. Compared with the current optimal method

it improves the recognition accuracy of scene ancient Chinese text by 3.27%.

Conclusions

In this paper

a multistage feature fusion network is designed to improve the feature extraction ability of the model

and the maximum boundary cosine loss is introduced to increase the distance between ancient Chinese text and narrow the distance within ancient Chinese text.

关键词

Keywords

references

MISHRA A ， ALAHARI K ， JAWAHAR C V . Image Retrieval Using Textual Cues ［C］// IEEE International Conference on Computer Vision ， Sydney， Australia ， 2014 .

韦豪东，易尧华，余长慧，等 . 融合注意力与序列单元的文本超分辨率［J］. 武汉大学学报（信息科学版）， 2024 ， 49 （ 7 ）： 1120 - 1129 .

WEI Haodong ， YI Yaohua ， YU Changhui ， et al . Text Super-Resolution Method with Attentional Mechanism and Sequential Units ［J］. Geomatics and Information Science of Wuhan University ， 2024 ， 49 （ 7 ）： 1120 - 1129 .

GOMEZ L ， MAFLA A ， RUSINOL M ， et al . Single Shot Scene Text Retrieval ［C］// Computer Vision–ECCV 2018 ， Cham， Switzerland ， 2018 .

YANG L J ， MA J ， XU T C ， et al . Fast Searching Chinese Calligraphic Information Based on Character Recognition ［C］// IEEE International Conference on Applied System Invention （ICASI）， Chiba， Japan ， 2018 .

HUANG J D ， CHENG G J ， ZHANG J H ， et al . Recognition Method for Stone Carved Calligraphy Characters Based on a Convolutional Neural Network ［J］. Neural Computing and Applications ， 2023 ， 35 （ 12 ）： 8723 - 8732 .

赵若晴，王慧琴，王可，等 . 基于方向梯度直方图和灰度共生矩阵混合特征的金文图像识别［J］. 激光与光电子学进展， 2020 ， 57 （ 12 ）： 98 - 104 .

ZHAO Ruoqing ， WANG Huiqin ， WANG Ke ， et al . Recognition of Bronze Inscriptions Image Based on Mixed Features of Histogram of Oriented Gradient and Gray Level Co-occurrence Matrix ［J］. Laser & Optoelectronics Progress ， 2020 ， 57 （ 12 ）： 98 - 104 .

陈丹，李宁，李亮 . 古文字的联机手写识别研究［J］. 北京机械工业学院学报， 2008 ， 23 （ 4 ）： 32 - 37 .

CHEN Dan ， LI Ning ， LI Liang . Online Recognition of Ancient Characters ［J］. Journal of Beijing Institute of Machinery ， 2008 ， 23 （ 4 ）： 32 - 37 .

SZEGEDY C ， LIU W ， JIA Y Q ， et al . Going Deeper with Convolutions ［C］// IEEE Conference on Computer Vision and Pattern Recognition （CVPR）， Boston， USA ， 2015 .

HE K M ， ZHANG X Y ， REN S Q ， et al . Deep Residual Learning for Image Recognition ［C］// 2016 IEEE Conference on Computer Vision and Pattern Recognition （CVPR）， Las Vegas， USA ， 2016 .

HUANG G ， LIU Z ， VAN DER MAATEN L ， et al . Densely Connected Convolutional Networks ［C］// 2017 IEEE Conference on Computer Vision and Pattern Recognition （CVPR）， Honolulu，USA ， 2017 .

CONG NGUYEN K ， TUAN NGUYEN C ， HOTTA S ， et al . A Character Attention Generative Adversarial Network for Degraded Historical Document Restoration ［C］// International Conference on Document Analysis and Recognition （ICDAR）， Sydney，Australia ， 2020 .

GOODFELLOW I ， POUGET A J ， MIRZA M ， et al . Generative Adversarial Nets ［J］. Advances in Neural Information Processing Systems ， 2014 ， 27 （ 3 ）： 54 - 68 .

MA W H ， ZHANG H S ， JIN L W ， et al . Joint Layout Analysis， Character Detection and Recognition for Historical Document Digitization ［C］// 17th International Conference on Frontiers in Handwriting Recognition （ICFHR）， Dortmund， Germany ， 2020 .

陈娅娅，刘全香，王凯丽，等 . 基于ResNet和迁移学习的古印章文本识别［J］. 计算机工程与应用， 2022 ， 58 （ 10 ）： 125 - 131 .

CHEN Yaya ， LIU Quanxiang ， WANG Kaili ， et al . Historical Chinese Seal Text Recognition Based on ResNet and Transfer Learning ［J］. Computer Engineering and Applications ， 2022 ， 58 （ 10 ）： 125 - 131 .

TANG M L ， XIE S M ， LIU X R . Ancient Character Recognition： A Novel Image Dataset of Shui Manuscript Characters and Classification Model ［J］. Chinese Journal of Electronics ， 2023 ， 32 （ 1 ）： 64 - 75 .

WANG K L ， YI Y H ， LIU J J ， et al . Multi-scene Ancient Chinese Text Recognition ［J］. Neurocomputing ， 2020 ， 377 ： 64 - 72 .

WANG K L ， YI Y H ， TANG Z W ， et al . Multi-scene Ancient Chinese Text Recognition with Deep Coupled Alignments ［J］. Applied Soft Computing ， 2021 ， 108 ： 107 - 125 .

时永欣，周维勋，邵振峰 . 融合多尺度注意力的多视角遥感影像场景分类［J］. 武汉大学学报（信息科学版）， 2024 ， 49 （ 3 ）： 366 - 375 .

SHI Yongxin ， ZHOU Weixun ， SHAO Zhenfeng . Multi-view Remote Sensing Image Scene Classification by Fusing Multi-Scale Attention ［J］. Geomatics and Information Science of Wuhan University ， 2024 ， 49 （ 3 ）： 366 - 375 .

YANG H ， YU H Y ， ZHENG K ， et al . Hyperspectral Image Classification Based on Interactive Transformer and CNN with Multilevel Feature Fusion Network ［J］. IEEE Geoscience and Remote Sensing Letters ， 2023 ， 20 ： 55 - 79 .

LIU S ， ZHANG Q ， HUANG L L . Graphic Image Classification Method Based on an Attention Mechanism and Fusion of Multilevel and Multiscale Deep Features ［J］. Computer Communications ， 2023 ， 209 ： 230 - 238 .

LI X ， WANG W H ， HU X L ， et al . Selective Kernel Networks ［C］// IEEE/CVF Conference on Computer Vision and Pattern Recognition （CVPR）， Long Beach， CA， USA， 2020 .

SZEGEDY C ， VANHOUCKE V ， IOFFE S ， et al . Rethinking the Inception Architecture for Computer Vision ［C］// IEEE Conference on Computer Vision and Pattern Recognition （CVPR）， Las Vegas， USA ， 2016 .

WANG H ， WANG Y T ， ZHOU Z ， et al . CosFace： Large Margin Cosine Loss for Deep Face Recognition ［C］// IEEE/CVF Conference on Computer Vision and Pattern Recognition ， Salt Lake City， USA ， 2018 .

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Spatiotemporal Intelligence： Theory and Applications

Development and Prospects of Spaceborne Greenhouse Gas LiDAR Technology

Research Progress in Eco-Geological Environmental Monitoring and Construction of an Intelligent Monitoring System in the Qinling Mountains

High-Resolution L-Band InSAR Study of Intense Near-Field Deformation Associated with Shallow Moderate Earthquakes

Intelligent Change Detection Method and Application Based on Wuhan-1 Satellite Imagery

Related Author

No data

Related Institution

No data

AI问答

⁰